A collaborative study by OpenAI and MIT Media Lab has revealed that intensive ChatGPT users are at risk of developing addiction-like behaviors. The study analyzed 40 million interactions and surveyed 1,000 participants over four weeks, identifying 'power users' as particularly vulnerable. The research found that addiction indicators include constant preoccupation, withdrawal symptoms, loss of control over usage duration, and mood changes tied to interactions with the AI. Users relying on ChatGPT for non-personal purposes, such as brainstorming or advice, are more likely to develop emotional dependency than those using it for personal discussions. Those experiencing loneliness or personal challenges are more prone to forming parasocial relationships with ChatGPT, often perceiving it as a 'friend.' The study also highlighted that text-based interactions with ChatGPT elicit more emotional language compared to its voice-based mode, which is linked to better short-term well-being. The research echoes earlier warnings from OpenAI regarding dependency risks associated with ChatGPT's audio features. It emphasizes the need to address the broader implications of prolonged generative AI usage, drawing parallels to dependency risks seen in social media and other digital platforms.